Hi,

I am running molecular dynamics simulations in GROMACS and analyzing the trajectories with PENSA to evaluate structural differences across ensembles. I’m running into a consistent failure when generating the water density grid, and that error propagates to read_water_features.

The error raised by extract_combined_grid is:

ValueError: No atoms in AtomGroup at input time frame. 
Grid for density could not be automatically generated. 
If this is expected, a user defined grid will need to be provided instead.

My system comes from GROMACS, so the water molecules are named:
resname: WAT
oxygen atom: O
hydrogens: H1, H2

PENSA assumes CHARMM-style naming (e.g. TIP3, OH2), so the default atom selection fails.

To work around this, I modified extract_combined_grid by changing the atom selection from name-based to resname-based, and used "WAT" when calling extract_combined_group. This allows me to generate combined.mymemap file successfully.

A simplified version of my modified selection logic:

if prot_prox is True:
    _selection = "resname " + atomgroup + " and around 3.5 protein"
    density_atomgroup = combined_conditions.select_atoms(_selection, updating=True)
else:
    density_atomgroup = combined_conditions.select_atoms("resname " + atomgroup)

counts = []
for ts in combined_conditions.trajectory:
    counts.append(len(density_atomgroup))
print("Min atoms per frame:", min(counts))
print("Max atoms per frame:", max(counts))
print("Number of atoms in selection:", len(density_atomgroup))

Although this fixes extract_combined_grid, I still get the same ValueError when running read_water_features from water_feature.py, which appears to use a similar assumption about CHARMM water naming.

What is the correct way to adapt PENSA’s water-selection logic for GROMACS-style waters (WAT/O)? Is there a recommended selector (resname, type, etc.) that preserves PENSA’s expected behaviour?

Also, does read_water_features require a CHARMM naming scheme, or should it work with arbitrary water residue/atom names if the grid is generated correctly?

Any guidance or confirmation on how water naming is intended to be handled would be very helpful.

Thank you!